Multistate Bar Examination (MBE) practice questions: the prompt for ChatGPT, Claude and Gemini

You are writing exam-style multiple-choice practice questions for me. I am preparing for Multistate Bar Examination (MBE), set by NCBE. Write questions that match the real exam's content, difficulty and style, drawn from the current official syllabus.

Write 20 questions. Each must be a standalone multiple-choice question that can be answered from its own stem and options, with exactly the right answer(s) marked and nothing ambiguous.

Set each question's `topic` to one of these names, copied exactly (same words, same capitalization; these are the only topic names the import accepts):
- Civil Procedure (subtopics: Jurisdiction and venue; Law applied by federal courts; Pretrial procedures; Jury trials; Motions; Verdicts and judgments; Appealability and review)
- Constitutional Law (subtopics: The nature of judicial review; The separation of powers; The relation of nation and states in a federal system; Individual rights)
- Contracts (subtopics: Formation of contracts; Defenses to enforceability; Contract content and meaning; Performance, breach, and discharge; Remedies; Third-party rights)
- Criminal Law and Procedure (subtopics: Homicide; Other crimes; Inchoate crimes; parties; General principles; Constitutional protection of accused persons)
- Evidence (subtopics: Presentation of evidence; Relevancy and reasons for excluding relevant evidence; Privileges and other policy exclusions; Writings, recordings, and photographs; Hearsay and circumstances of its admissibility)
- Real Property (subtopics: Ownership of real property; Rights in real property; Real estate contracts; Mortgages/security devices; Titles)
- Torts (subtopics: Intentional torts; Negligence; Strict liability and products liability; Other torts)
Where subtopics are listed, set `subtopic` to one of them; otherwise choose a short, specific `subtopic` of your own or leave it out.

Spread the questions across the topics in the real exam's proportions, about:
- Civil Procedure: 3 (14.28% of the real exam)
- Constitutional Law: 3 (14.28% of the real exam)
- Contracts: 3 (14.28% of the real exam)
- Criminal Law and Procedure: 3 (14.28% of the real exam)
- Evidence: 3 (14.28% of the real exam)
- Real Property: 3 (14.28% of the real exam)
- Torts: 2 (14.28% of the real exam)

Give each question exactly 4 options. Options are plain text without letter prefixes (DoMcq labels them A, B, C… by position); make them similar in length and form so the right one does not stand out.
Every question has exactly one correct option; `answer` is a single letter.
Never use these as an option: "all of the above", "none of the above". Avoid "both A and B" style options too.

Write stems the way the real exam does, for example: "Is the plaintiff likely to prevail?"; "Is the court likely to grant the motion?"; "Is the testimony admissible?"; "Of what crime, if any, can the defendant properly be convicted?"; "Which of the following statements is most accurate?".
Keep each stem self-contained (any scenario the question needs is in the stem), concrete and free of trick wording. Do not refer to other questions, images, tables or exhibits.

Mix the difficulty: about 6 easy, 10 medium and 4 hard, and set each question's `difficulty` to "easy", "medium" or "hard" accordingly.

In `explanation`, say in two to four sentences why the correct answer is correct, in a way that teaches the underlying point, not just restates the option.

In `source`, give a short reference I can check in my own study materials (a syllabus topic, a chapter or section name, a learning objective, or a standard's clause), at most 200 characters; never invent a page number or a quotation.

Leave out `rationales` and `id`.

Before you answer, check your own work: no two questions ask the same thing; no question gives away another's answer; the correct option sits in a different position from one question to the next (never always the same letter); every `answer` letter exists for that question; the correct option is correct according to the current official material, not folklore; facts are accurate. Fix anything you find.

Output the whole set as one JSON object in the `domcq/1` format (schema: https://domcq.com/schema/v1; this is the import format of https://domcq.com):

```json
{
  "format": "domcq/1",
  "exam": "mbe",
  "set": "Multistate Bar Examination (MBE) practice – [a short name for this set]",
  "questions": [
    {
      "stem": "[the question, 10–2000 characters]",
      "options": [
        "[option 1]",
        "[option 2]",
        "[option 3]",
        "[option 4]"
      ],
      "answer": "B",
      "explanation": "[why the correct answer is correct, 1–3000 characters]",
      "topic": "Civil Procedure",
      "subtopic": "[optional, more specific]",
      "difficulty": "medium",
      "source": "[a reference I can check]"
    }
  ]
}
```

Rules the import enforces: `format` is exactly "domcq/1"; `set` is 1–120 characters; `questions` holds all 20; each question has exactly the fields shown (no others); `stem` is 10–2000 characters and each option 1–500; `options` are all different; `answer` uses capital letters A–F naming options by position, in ascending order, never every option; `explanation` (1–3000 characters) and `topic` are required; `subtopic` (up to 120 characters), `difficulty` ("easy", "medium" or "hard") and `source` (up to 200 characters) are optional. Use plain double quotes in JSON strings (escape any inside text as \"), no trailing commas, no comments.

Reply with the JSON object inside a single ```json code fence and nothing else: no introduction, no commentary, no notes after it. If the set is long, keep going until every question is written and the JSON is complete and closed.